On Wednesday, January 22, 2003, at 12:57  PM, Yukihiro Matsumoto wrote:

> Hi,
>
> In message "Re: Extensions for Time and ParseDate"
>     on 03/01/23, Mathieu Bouchard <matju / sympatico.ca> writes:
>
> |> Ryan, I also fixed EWOULDBLOCK problem.  I will commit them soon.
> |
> |May I know what that problem is? I'm curious, because two of my
> |libraries depend on nonblocking I/O (GridFlow, RubyX11).
>
> 1.6.8 and 1.8.0 does not provide duplicated error constants, so that
> when EAGAIN == EWOULDBLOCK, there's no Errno::EWOULDBLOCK defined.

I think it's simpler than that... In the case of the header I was 
looking at:

#ifndef _POSIX_SOURCE
#define EWOULDBLOCK EAGAIN
...
#endif

but ruby is compiled w/ _POSIX_SOURCE defined (correctly) so 
EWOULDBLOCK shouldn't be defined. I think we were just dealing w/ code 
that was poorly written (overlooked the #ifndef) or wasn't in fact 
posix compliant.